root / lib / cmdlib / node.py @ 4869595d
History | View | Annotate | Download (56.9 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc.
|
5 |
#
|
6 |
# This program is free software; you can redistribute it and/or modify
|
7 |
# it under the terms of the GNU General Public License as published by
|
8 |
# the Free Software Foundation; either version 2 of the License, or
|
9 |
# (at your option) any later version.
|
10 |
#
|
11 |
# This program is distributed in the hope that it will be useful, but
|
12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
14 |
# General Public License for more details.
|
15 |
#
|
16 |
# You should have received a copy of the GNU General Public License
|
17 |
# along with this program; if not, write to the Free Software
|
18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
19 |
# 02110-1301, USA.
|
20 |
|
21 |
|
22 |
"""Logical units dealing with nodes."""
|
23 |
|
24 |
import logging |
25 |
import operator |
26 |
|
27 |
from ganeti import constants |
28 |
from ganeti import errors |
29 |
from ganeti import locking |
30 |
from ganeti import netutils |
31 |
from ganeti import objects |
32 |
from ganeti import opcodes |
33 |
import ganeti.rpc.node as rpc |
34 |
from ganeti import utils |
35 |
from ganeti.masterd import iallocator |
36 |
|
37 |
from ganeti.cmdlib.base import LogicalUnit, NoHooksLU, ResultWithJobs |
38 |
from ganeti.cmdlib.common import CheckParamsNotGlobal, \ |
39 |
MergeAndVerifyHvState, MergeAndVerifyDiskState, \ |
40 |
IsExclusiveStorageEnabledNode, CheckNodePVs, \ |
41 |
RedistributeAncillaryFiles, ExpandNodeUuidAndName, ShareAll, SupportsOob, \ |
42 |
CheckInstanceState, INSTANCE_DOWN, GetUpdatedParams, \ |
43 |
AdjustCandidatePool, CheckIAllocatorOrNode, LoadNodeEvacResult, \ |
44 |
GetWantedNodes, MapInstanceLvsToNodes, RunPostHook, \ |
45 |
FindFaultyInstanceDisks, CheckStorageTypeEnabled |
46 |
|
47 |
|
48 |
def _DecideSelfPromotion(lu, exceptions=None): |
49 |
"""Decide whether I should promote myself as a master candidate.
|
50 |
|
51 |
"""
|
52 |
cp_size = lu.cfg.GetClusterInfo().candidate_pool_size |
53 |
mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions) |
54 |
# the new node will increase mc_max with one, so:
|
55 |
mc_should = min(mc_should + 1, cp_size) |
56 |
return mc_now < mc_should
|
57 |
|
58 |
|
59 |
def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq): |
60 |
"""Ensure that a node has the given secondary ip.
|
61 |
|
62 |
@type lu: L{LogicalUnit}
|
63 |
@param lu: the LU on behalf of which we make the check
|
64 |
@type node: L{objects.Node}
|
65 |
@param node: the node to check
|
66 |
@type secondary_ip: string
|
67 |
@param secondary_ip: the ip to check
|
68 |
@type prereq: boolean
|
69 |
@param prereq: whether to throw a prerequisite or an execute error
|
70 |
@raise errors.OpPrereqError: if the node doesn't have the ip,
|
71 |
and prereq=True
|
72 |
@raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
|
73 |
|
74 |
"""
|
75 |
# this can be called with a new node, which has no UUID yet, so perform the
|
76 |
# RPC call using its name
|
77 |
result = lu.rpc.call_node_has_ip_address(node.name, secondary_ip) |
78 |
result.Raise("Failure checking secondary ip on node %s" % node.name,
|
79 |
prereq=prereq, ecode=errors.ECODE_ENVIRON) |
80 |
if not result.payload: |
81 |
msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
|
82 |
" please fix and re-run this command" % secondary_ip)
|
83 |
if prereq:
|
84 |
raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
|
85 |
else:
|
86 |
raise errors.OpExecError(msg)
|
87 |
|
88 |
|
89 |
class LUNodeAdd(LogicalUnit): |
90 |
"""Logical unit for adding node to the cluster.
|
91 |
|
92 |
"""
|
93 |
HPATH = "node-add"
|
94 |
HTYPE = constants.HTYPE_NODE |
95 |
_NFLAGS = ["master_capable", "vm_capable"] |
96 |
|
97 |
def CheckArguments(self): |
98 |
self.primary_ip_family = self.cfg.GetPrimaryIPFamily() |
99 |
# validate/normalize the node name
|
100 |
self.hostname = netutils.GetHostname(name=self.op.node_name, |
101 |
family=self.primary_ip_family)
|
102 |
self.op.node_name = self.hostname.name |
103 |
|
104 |
if self.op.readd and self.op.node_name == self.cfg.GetMasterNodeName(): |
105 |
raise errors.OpPrereqError("Cannot readd the master node", |
106 |
errors.ECODE_STATE) |
107 |
|
108 |
if self.op.readd and self.op.group: |
109 |
raise errors.OpPrereqError("Cannot pass a node group when a node is" |
110 |
" being readded", errors.ECODE_INVAL)
|
111 |
|
112 |
def BuildHooksEnv(self): |
113 |
"""Build hooks env.
|
114 |
|
115 |
This will run on all nodes before, and on all nodes + the new node after.
|
116 |
|
117 |
"""
|
118 |
return {
|
119 |
"OP_TARGET": self.op.node_name, |
120 |
"NODE_NAME": self.op.node_name, |
121 |
"NODE_PIP": self.op.primary_ip, |
122 |
"NODE_SIP": self.op.secondary_ip, |
123 |
"MASTER_CAPABLE": str(self.op.master_capable), |
124 |
"VM_CAPABLE": str(self.op.vm_capable), |
125 |
} |
126 |
|
127 |
def BuildHooksNodes(self): |
128 |
"""Build hooks nodes.
|
129 |
|
130 |
"""
|
131 |
hook_nodes = self.cfg.GetNodeList()
|
132 |
new_node_info = self.cfg.GetNodeInfoByName(self.op.node_name) |
133 |
if new_node_info is not None: |
134 |
# Exclude added node
|
135 |
hook_nodes = list(set(hook_nodes) - set([new_node_info.uuid])) |
136 |
|
137 |
# add the new node as post hook node by name; it does not have an UUID yet
|
138 |
return (hook_nodes, hook_nodes, [self.op.node_name, ]) |
139 |
|
140 |
def CheckPrereq(self): |
141 |
"""Check prerequisites.
|
142 |
|
143 |
This checks:
|
144 |
- the new node is not already in the config
|
145 |
- it is resolvable
|
146 |
- its parameters (single/dual homed) matches the cluster
|
147 |
|
148 |
Any errors are signaled by raising errors.OpPrereqError.
|
149 |
|
150 |
"""
|
151 |
node_name = self.hostname.name
|
152 |
self.op.primary_ip = self.hostname.ip |
153 |
if self.op.secondary_ip is None: |
154 |
if self.primary_ip_family == netutils.IP6Address.family: |
155 |
raise errors.OpPrereqError("When using a IPv6 primary address, a valid" |
156 |
" IPv4 address must be given as secondary",
|
157 |
errors.ECODE_INVAL) |
158 |
self.op.secondary_ip = self.op.primary_ip |
159 |
|
160 |
secondary_ip = self.op.secondary_ip
|
161 |
if not netutils.IP4Address.IsValid(secondary_ip): |
162 |
raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4" |
163 |
" address" % secondary_ip, errors.ECODE_INVAL)
|
164 |
|
165 |
existing_node_info = self.cfg.GetNodeInfoByName(node_name)
|
166 |
if not self.op.readd and existing_node_info is not None: |
167 |
raise errors.OpPrereqError("Node %s is already in the configuration" % |
168 |
node_name, errors.ECODE_EXISTS) |
169 |
elif self.op.readd and existing_node_info is None: |
170 |
raise errors.OpPrereqError("Node %s is not in the configuration" % |
171 |
node_name, errors.ECODE_NOENT) |
172 |
|
173 |
self.changed_primary_ip = False |
174 |
|
175 |
for existing_node in self.cfg.GetAllNodesInfo().values(): |
176 |
if self.op.readd and node_name == existing_node.name: |
177 |
if existing_node.secondary_ip != secondary_ip:
|
178 |
raise errors.OpPrereqError("Readded node doesn't have the same IP" |
179 |
" address configuration as before",
|
180 |
errors.ECODE_INVAL) |
181 |
if existing_node.primary_ip != self.op.primary_ip: |
182 |
self.changed_primary_ip = True |
183 |
|
184 |
continue
|
185 |
|
186 |
if (existing_node.primary_ip == self.op.primary_ip or |
187 |
existing_node.secondary_ip == self.op.primary_ip or |
188 |
existing_node.primary_ip == secondary_ip or
|
189 |
existing_node.secondary_ip == secondary_ip): |
190 |
raise errors.OpPrereqError("New node ip address(es) conflict with" |
191 |
" existing node %s" % existing_node.name,
|
192 |
errors.ECODE_NOTUNIQUE) |
193 |
|
194 |
# After this 'if' block, None is no longer a valid value for the
|
195 |
# _capable op attributes
|
196 |
if self.op.readd: |
197 |
assert existing_node_info is not None, \ |
198 |
"Can't retrieve locked node %s" % node_name
|
199 |
for attr in self._NFLAGS: |
200 |
if getattr(self.op, attr) is None: |
201 |
setattr(self.op, attr, getattr(existing_node_info, attr)) |
202 |
else:
|
203 |
for attr in self._NFLAGS: |
204 |
if getattr(self.op, attr) is None: |
205 |
setattr(self.op, attr, True) |
206 |
|
207 |
if self.op.readd and not self.op.vm_capable: |
208 |
pri, sec = self.cfg.GetNodeInstances(existing_node_info.uuid)
|
209 |
if pri or sec: |
210 |
raise errors.OpPrereqError("Node %s being re-added with vm_capable" |
211 |
" flag set to false, but it already holds"
|
212 |
" instances" % node_name,
|
213 |
errors.ECODE_STATE) |
214 |
|
215 |
# check that the type of the node (single versus dual homed) is the
|
216 |
# same as for the master
|
217 |
myself = self.cfg.GetMasterNodeInfo()
|
218 |
master_singlehomed = myself.secondary_ip == myself.primary_ip |
219 |
newbie_singlehomed = secondary_ip == self.op.primary_ip
|
220 |
if master_singlehomed != newbie_singlehomed:
|
221 |
if master_singlehomed:
|
222 |
raise errors.OpPrereqError("The master has no secondary ip but the" |
223 |
" new node has one",
|
224 |
errors.ECODE_INVAL) |
225 |
else:
|
226 |
raise errors.OpPrereqError("The master has a secondary ip but the" |
227 |
" new node doesn't have one",
|
228 |
errors.ECODE_INVAL) |
229 |
|
230 |
# checks reachability
|
231 |
if not netutils.TcpPing(self.op.primary_ip, constants.DEFAULT_NODED_PORT): |
232 |
raise errors.OpPrereqError("Node not reachable by ping", |
233 |
errors.ECODE_ENVIRON) |
234 |
|
235 |
if not newbie_singlehomed: |
236 |
# check reachability from my secondary ip to newbie's secondary ip
|
237 |
if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT, |
238 |
source=myself.secondary_ip): |
239 |
raise errors.OpPrereqError("Node secondary ip not reachable by TCP" |
240 |
" based ping to node daemon port",
|
241 |
errors.ECODE_ENVIRON) |
242 |
|
243 |
if self.op.readd: |
244 |
exceptions = [existing_node_info.uuid] |
245 |
else:
|
246 |
exceptions = [] |
247 |
|
248 |
if self.op.master_capable: |
249 |
self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions) |
250 |
else:
|
251 |
self.master_candidate = False |
252 |
|
253 |
node_group = self.cfg.LookupNodeGroup(self.op.group) |
254 |
|
255 |
if self.op.readd: |
256 |
self.new_node = existing_node_info
|
257 |
else:
|
258 |
self.new_node = objects.Node(name=node_name,
|
259 |
primary_ip=self.op.primary_ip,
|
260 |
secondary_ip=secondary_ip, |
261 |
master_candidate=self.master_candidate,
|
262 |
offline=False, drained=False, |
263 |
group=node_group, ndparams={}) |
264 |
|
265 |
if self.op.ndparams: |
266 |
utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
|
267 |
CheckParamsNotGlobal(self.op.ndparams, constants.NDC_GLOBALS, "node", |
268 |
"node", "cluster or group") |
269 |
|
270 |
if self.op.hv_state: |
271 |
self.new_hv_state = MergeAndVerifyHvState(self.op.hv_state, None) |
272 |
|
273 |
if self.op.disk_state: |
274 |
self.new_disk_state = MergeAndVerifyDiskState(self.op.disk_state, None) |
275 |
|
276 |
# TODO: If we need to have multiple DnsOnlyRunner we probably should make
|
277 |
# it a property on the base class.
|
278 |
rpcrunner = rpc.DnsOnlyRunner() |
279 |
result = rpcrunner.call_version([node_name])[node_name] |
280 |
result.Raise("Can't get version information from node %s" % node_name)
|
281 |
if constants.PROTOCOL_VERSION == result.payload:
|
282 |
logging.info("Communication to node %s fine, sw version %s match",
|
283 |
node_name, result.payload) |
284 |
else:
|
285 |
raise errors.OpPrereqError("Version mismatch master version %s," |
286 |
" node version %s" %
|
287 |
(constants.PROTOCOL_VERSION, result.payload), |
288 |
errors.ECODE_ENVIRON) |
289 |
|
290 |
vg_name = self.cfg.GetVGName()
|
291 |
if vg_name is not None: |
292 |
vparams = {constants.NV_PVLIST: [vg_name]} |
293 |
excl_stor = IsExclusiveStorageEnabledNode(self.cfg, self.new_node) |
294 |
cname = self.cfg.GetClusterName()
|
295 |
result = rpcrunner.call_node_verify_light( |
296 |
[node_name], vparams, cname, |
297 |
self.cfg.GetClusterInfo().hvparams,
|
298 |
{node_name: node_group}, |
299 |
self.cfg.GetAllNodeGroupsInfoDict()
|
300 |
)[node_name] |
301 |
(errmsgs, _) = CheckNodePVs(result.payload, excl_stor) |
302 |
if errmsgs:
|
303 |
raise errors.OpPrereqError("Checks on node PVs failed: %s" % |
304 |
"; ".join(errmsgs), errors.ECODE_ENVIRON)
|
305 |
|
306 |
def _InitOpenVSwitch(self): |
307 |
filled_ndparams = self.cfg.GetClusterInfo().FillND(
|
308 |
self.new_node, self.cfg.GetNodeGroup(self.new_node.group)) |
309 |
|
310 |
ovs = filled_ndparams.get(constants.ND_OVS, None)
|
311 |
ovs_name = filled_ndparams.get(constants.ND_OVS_NAME, None)
|
312 |
ovs_link = filled_ndparams.get(constants.ND_OVS_LINK, None)
|
313 |
|
314 |
if ovs:
|
315 |
if not ovs_link: |
316 |
self.LogInfo("No physical interface for OpenvSwitch was given." |
317 |
" OpenvSwitch will not have an outside connection. This"
|
318 |
" might not be what you want.")
|
319 |
|
320 |
result = self.rpc.call_node_configure_ovs(
|
321 |
self.new_node.name, ovs_name, ovs_link)
|
322 |
result.Raise("Failed to initialize OpenVSwitch on new node")
|
323 |
|
324 |
def Exec(self, feedback_fn): |
325 |
"""Adds the new node to the cluster.
|
326 |
|
327 |
"""
|
328 |
assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \ |
329 |
"Not owning BGL"
|
330 |
|
331 |
# We adding a new node so we assume it's powered
|
332 |
self.new_node.powered = True |
333 |
|
334 |
# for re-adds, reset the offline/drained/master-candidate flags;
|
335 |
# we need to reset here, otherwise offline would prevent RPC calls
|
336 |
# later in the procedure; this also means that if the re-add
|
337 |
# fails, we are left with a non-offlined, broken node
|
338 |
if self.op.readd: |
339 |
self.new_node.offline = False |
340 |
self.new_node.drained = False |
341 |
self.LogInfo("Readding a node, the offline/drained flags were reset") |
342 |
# if we demote the node, we do cleanup later in the procedure
|
343 |
self.new_node.master_candidate = self.master_candidate |
344 |
if self.changed_primary_ip: |
345 |
self.new_node.primary_ip = self.op.primary_ip |
346 |
|
347 |
# copy the master/vm_capable flags
|
348 |
for attr in self._NFLAGS: |
349 |
setattr(self.new_node, attr, getattr(self.op, attr)) |
350 |
|
351 |
# notify the user about any possible mc promotion
|
352 |
if self.new_node.master_candidate: |
353 |
self.LogInfo("Node will be a master candidate") |
354 |
|
355 |
if self.op.ndparams: |
356 |
self.new_node.ndparams = self.op.ndparams |
357 |
else:
|
358 |
self.new_node.ndparams = {}
|
359 |
|
360 |
if self.op.hv_state: |
361 |
self.new_node.hv_state_static = self.new_hv_state |
362 |
|
363 |
if self.op.disk_state: |
364 |
self.new_node.disk_state_static = self.new_disk_state |
365 |
|
366 |
# Add node to our /etc/hosts, and add key to known_hosts
|
367 |
if self.cfg.GetClusterInfo().modify_etc_hosts: |
368 |
master_node = self.cfg.GetMasterNode()
|
369 |
result = self.rpc.call_etc_hosts_modify(
|
370 |
master_node, constants.ETC_HOSTS_ADD, self.hostname.name,
|
371 |
self.hostname.ip)
|
372 |
result.Raise("Can't update hosts file with new host data")
|
373 |
|
374 |
if self.new_node.secondary_ip != self.new_node.primary_ip: |
375 |
_CheckNodeHasSecondaryIP(self, self.new_node, self.new_node.secondary_ip, |
376 |
False)
|
377 |
|
378 |
node_verifier_uuids = [self.cfg.GetMasterNode()]
|
379 |
node_verify_param = { |
380 |
constants.NV_NODELIST: ([self.new_node.name], {}),
|
381 |
# TODO: do a node-net-test as well?
|
382 |
} |
383 |
|
384 |
result = self.rpc.call_node_verify(
|
385 |
node_verifier_uuids, node_verify_param, |
386 |
self.cfg.GetClusterName(),
|
387 |
self.cfg.GetClusterInfo().hvparams,
|
388 |
{self.new_node.name: self.cfg.LookupNodeGroup(self.op.group)}, |
389 |
self.cfg.GetAllNodeGroupsInfoDict()
|
390 |
) |
391 |
for verifier in node_verifier_uuids: |
392 |
result[verifier].Raise("Cannot communicate with node %s" % verifier)
|
393 |
nl_payload = result[verifier].payload[constants.NV_NODELIST] |
394 |
if nl_payload:
|
395 |
for failed in nl_payload: |
396 |
feedback_fn("ssh/hostname verification failed"
|
397 |
" (checking from %s): %s" %
|
398 |
(verifier, nl_payload[failed])) |
399 |
raise errors.OpExecError("ssh/hostname verification failed") |
400 |
|
401 |
self._InitOpenVSwitch()
|
402 |
|
403 |
if self.op.readd: |
404 |
self.context.ReaddNode(self.new_node) |
405 |
RedistributeAncillaryFiles(self)
|
406 |
# make sure we redistribute the config
|
407 |
self.cfg.Update(self.new_node, feedback_fn) |
408 |
# and make sure the new node will not have old files around
|
409 |
if not self.new_node.master_candidate: |
410 |
result = self.rpc.call_node_demote_from_mc(self.new_node.uuid) |
411 |
result.Warn("Node failed to demote itself from master candidate status",
|
412 |
self.LogWarning)
|
413 |
else:
|
414 |
self.context.AddNode(self.new_node, self.proc.GetECId()) |
415 |
RedistributeAncillaryFiles(self)
|
416 |
|
417 |
|
418 |
class LUNodeSetParams(LogicalUnit): |
419 |
"""Modifies the parameters of a node.
|
420 |
|
421 |
@cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
|
422 |
to the node role (as _ROLE_*)
|
423 |
@cvar _R2F: a dictionary from node role to tuples of flags
|
424 |
@cvar _FLAGS: a list of attribute names corresponding to the flags
|
425 |
|
426 |
"""
|
427 |
HPATH = "node-modify"
|
428 |
HTYPE = constants.HTYPE_NODE |
429 |
REQ_BGL = False
|
430 |
(_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4) |
431 |
_F2R = { |
432 |
(True, False, False): _ROLE_CANDIDATE, |
433 |
(False, True, False): _ROLE_DRAINED, |
434 |
(False, False, True): _ROLE_OFFLINE, |
435 |
(False, False, False): _ROLE_REGULAR, |
436 |
} |
437 |
_R2F = dict((v, k) for k, v in _F2R.items()) |
438 |
_FLAGS = ["master_candidate", "drained", "offline"] |
439 |
|
440 |
def CheckArguments(self): |
441 |
(self.op.node_uuid, self.op.node_name) = \ |
442 |
ExpandNodeUuidAndName(self.cfg, self.op.node_uuid, self.op.node_name) |
443 |
all_mods = [self.op.offline, self.op.master_candidate, self.op.drained, |
444 |
self.op.master_capable, self.op.vm_capable, |
445 |
self.op.secondary_ip, self.op.ndparams, self.op.hv_state, |
446 |
self.op.disk_state]
|
447 |
if all_mods.count(None) == len(all_mods): |
448 |
raise errors.OpPrereqError("Please pass at least one modification", |
449 |
errors.ECODE_INVAL) |
450 |
if all_mods.count(True) > 1: |
451 |
raise errors.OpPrereqError("Can't set the node into more than one" |
452 |
" state at the same time",
|
453 |
errors.ECODE_INVAL) |
454 |
|
455 |
# Boolean value that tells us whether we might be demoting from MC
|
456 |
self.might_demote = (self.op.master_candidate is False or |
457 |
self.op.offline is True or |
458 |
self.op.drained is True or |
459 |
self.op.master_capable is False) |
460 |
|
461 |
if self.op.secondary_ip: |
462 |
if not netutils.IP4Address.IsValid(self.op.secondary_ip): |
463 |
raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4" |
464 |
" address" % self.op.secondary_ip, |
465 |
errors.ECODE_INVAL) |
466 |
|
467 |
self.lock_all = self.op.auto_promote and self.might_demote |
468 |
self.lock_instances = self.op.secondary_ip is not None |
469 |
|
470 |
def _InstanceFilter(self, instance): |
471 |
"""Filter for getting affected instances.
|
472 |
|
473 |
"""
|
474 |
return (instance.disk_template in constants.DTS_INT_MIRROR and |
475 |
self.op.node_uuid in instance.all_nodes) |
476 |
|
477 |
def ExpandNames(self): |
478 |
if self.lock_all: |
479 |
self.needed_locks = {
|
480 |
locking.LEVEL_NODE: locking.ALL_SET, |
481 |
|
482 |
# Block allocations when all nodes are locked
|
483 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
484 |
} |
485 |
else:
|
486 |
self.needed_locks = {
|
487 |
locking.LEVEL_NODE: self.op.node_uuid,
|
488 |
} |
489 |
|
490 |
# Since modifying a node can have severe effects on currently running
|
491 |
# operations the resource lock is at least acquired in shared mode
|
492 |
self.needed_locks[locking.LEVEL_NODE_RES] = \
|
493 |
self.needed_locks[locking.LEVEL_NODE]
|
494 |
|
495 |
# Get all locks except nodes in shared mode; they are not used for anything
|
496 |
# but read-only access
|
497 |
self.share_locks = ShareAll()
|
498 |
self.share_locks[locking.LEVEL_NODE] = 0 |
499 |
self.share_locks[locking.LEVEL_NODE_RES] = 0 |
500 |
self.share_locks[locking.LEVEL_NODE_ALLOC] = 0 |
501 |
|
502 |
if self.lock_instances: |
503 |
self.needed_locks[locking.LEVEL_INSTANCE] = \
|
504 |
self.cfg.GetInstanceNames(
|
505 |
self.cfg.GetInstancesInfoByFilter(self._InstanceFilter).keys()) |
506 |
|
507 |
def BuildHooksEnv(self): |
508 |
"""Build hooks env.
|
509 |
|
510 |
This runs on the master node.
|
511 |
|
512 |
"""
|
513 |
return {
|
514 |
"OP_TARGET": self.op.node_name, |
515 |
"MASTER_CANDIDATE": str(self.op.master_candidate), |
516 |
"OFFLINE": str(self.op.offline), |
517 |
"DRAINED": str(self.op.drained), |
518 |
"MASTER_CAPABLE": str(self.op.master_capable), |
519 |
"VM_CAPABLE": str(self.op.vm_capable), |
520 |
} |
521 |
|
522 |
def BuildHooksNodes(self): |
523 |
"""Build hooks nodes.
|
524 |
|
525 |
"""
|
526 |
nl = [self.cfg.GetMasterNode(), self.op.node_uuid] |
527 |
return (nl, nl)
|
528 |
|
529 |
def CheckPrereq(self): |
530 |
"""Check prerequisites.
|
531 |
|
532 |
This only checks the instance list against the existing names.
|
533 |
|
534 |
"""
|
535 |
node = self.cfg.GetNodeInfo(self.op.node_uuid) |
536 |
if self.lock_instances: |
537 |
affected_instances = \ |
538 |
self.cfg.GetInstancesInfoByFilter(self._InstanceFilter) |
539 |
|
540 |
# Verify instance locks
|
541 |
owned_instance_names = self.owned_locks(locking.LEVEL_INSTANCE)
|
542 |
wanted_instance_names = frozenset([inst.name for inst in |
543 |
affected_instances.values()]) |
544 |
if wanted_instance_names - owned_instance_names:
|
545 |
raise errors.OpPrereqError("Instances affected by changing node %s's" |
546 |
" secondary IP address have changed since"
|
547 |
" locks were acquired, wanted '%s', have"
|
548 |
" '%s'; retry the operation" %
|
549 |
(node.name, |
550 |
utils.CommaJoin(wanted_instance_names), |
551 |
utils.CommaJoin(owned_instance_names)), |
552 |
errors.ECODE_STATE) |
553 |
else:
|
554 |
affected_instances = None
|
555 |
|
556 |
if (self.op.master_candidate is not None or |
557 |
self.op.drained is not None or |
558 |
self.op.offline is not None): |
559 |
# we can't change the master's node flags
|
560 |
if node.uuid == self.cfg.GetMasterNode(): |
561 |
raise errors.OpPrereqError("The master role can be changed" |
562 |
" only via master-failover",
|
563 |
errors.ECODE_INVAL) |
564 |
|
565 |
if self.op.master_candidate and not node.master_capable: |
566 |
raise errors.OpPrereqError("Node %s is not master capable, cannot make" |
567 |
" it a master candidate" % node.name,
|
568 |
errors.ECODE_STATE) |
569 |
|
570 |
if self.op.vm_capable is False: |
571 |
(ipri, isec) = self.cfg.GetNodeInstances(node.uuid)
|
572 |
if ipri or isec: |
573 |
raise errors.OpPrereqError("Node %s hosts instances, cannot unset" |
574 |
" the vm_capable flag" % node.name,
|
575 |
errors.ECODE_STATE) |
576 |
|
577 |
if node.master_candidate and self.might_demote and not self.lock_all: |
578 |
assert not self.op.auto_promote, "auto_promote set but lock_all not" |
579 |
# check if after removing the current node, we're missing master
|
580 |
# candidates
|
581 |
(mc_remaining, mc_should, _) = \ |
582 |
self.cfg.GetMasterCandidateStats(exceptions=[node.uuid])
|
583 |
if mc_remaining < mc_should:
|
584 |
raise errors.OpPrereqError("Not enough master candidates, please" |
585 |
" pass auto promote option to allow"
|
586 |
" promotion (--auto-promote or RAPI"
|
587 |
" auto_promote=True)", errors.ECODE_STATE)
|
588 |
|
589 |
self.old_flags = old_flags = (node.master_candidate,
|
590 |
node.drained, node.offline) |
591 |
assert old_flags in self._F2R, "Un-handled old flags %s" % str(old_flags) |
592 |
self.old_role = old_role = self._F2R[old_flags] |
593 |
|
594 |
# Check for ineffective changes
|
595 |
for attr in self._FLAGS: |
596 |
if getattr(self.op, attr) is False and getattr(node, attr) is False: |
597 |
self.LogInfo("Ignoring request to unset flag %s, already unset", attr) |
598 |
setattr(self.op, attr, None) |
599 |
|
600 |
# Past this point, any flag change to False means a transition
|
601 |
# away from the respective state, as only real changes are kept
|
602 |
|
603 |
# TODO: We might query the real power state if it supports OOB
|
604 |
if SupportsOob(self.cfg, node): |
605 |
if self.op.offline is False and not (node.powered or |
606 |
self.op.powered is True): |
607 |
raise errors.OpPrereqError(("Node %s needs to be turned on before its" |
608 |
" offline status can be reset") %
|
609 |
self.op.node_name, errors.ECODE_STATE)
|
610 |
elif self.op.powered is not None: |
611 |
raise errors.OpPrereqError(("Unable to change powered state for node %s" |
612 |
" as it does not support out-of-band"
|
613 |
" handling") % self.op.node_name, |
614 |
errors.ECODE_STATE) |
615 |
|
616 |
# If we're being deofflined/drained, we'll MC ourself if needed
|
617 |
if (self.op.drained is False or self.op.offline is False or |
618 |
(self.op.master_capable and not node.master_capable)): |
619 |
if _DecideSelfPromotion(self): |
620 |
self.op.master_candidate = True |
621 |
self.LogInfo("Auto-promoting node to master candidate") |
622 |
|
623 |
# If we're no longer master capable, we'll demote ourselves from MC
|
624 |
if self.op.master_capable is False and node.master_candidate: |
625 |
self.LogInfo("Demoting from master candidate") |
626 |
self.op.master_candidate = False |
627 |
|
628 |
# Compute new role
|
629 |
assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1 |
630 |
if self.op.master_candidate: |
631 |
new_role = self._ROLE_CANDIDATE
|
632 |
elif self.op.drained: |
633 |
new_role = self._ROLE_DRAINED
|
634 |
elif self.op.offline: |
635 |
new_role = self._ROLE_OFFLINE
|
636 |
elif False in [self.op.master_candidate, self.op.drained, self.op.offline]: |
637 |
# False is still in new flags, which means we're un-setting (the
|
638 |
# only) True flag
|
639 |
new_role = self._ROLE_REGULAR
|
640 |
else: # no new flags, nothing, keep old role |
641 |
new_role = old_role |
642 |
|
643 |
self.new_role = new_role
|
644 |
|
645 |
if old_role == self._ROLE_OFFLINE and new_role != old_role: |
646 |
# Trying to transition out of offline status
|
647 |
result = self.rpc.call_version([node.uuid])[node.uuid]
|
648 |
if result.fail_msg:
|
649 |
raise errors.OpPrereqError("Node %s is being de-offlined but fails" |
650 |
" to report its version: %s" %
|
651 |
(node.name, result.fail_msg), |
652 |
errors.ECODE_STATE) |
653 |
else:
|
654 |
self.LogWarning("Transitioning node from offline to online state" |
655 |
" without using re-add. Please make sure the node"
|
656 |
" is healthy!")
|
657 |
|
658 |
# When changing the secondary ip, verify if this is a single-homed to
|
659 |
# multi-homed transition or vice versa, and apply the relevant
|
660 |
# restrictions.
|
661 |
if self.op.secondary_ip: |
662 |
# Ok even without locking, because this can't be changed by any LU
|
663 |
master = self.cfg.GetMasterNodeInfo()
|
664 |
master_singlehomed = master.secondary_ip == master.primary_ip |
665 |
if master_singlehomed and self.op.secondary_ip != node.primary_ip: |
666 |
if self.op.force and node.uuid == master.uuid: |
667 |
self.LogWarning("Transitioning from single-homed to multi-homed" |
668 |
" cluster; all nodes will require a secondary IP"
|
669 |
" address")
|
670 |
else:
|
671 |
raise errors.OpPrereqError("Changing the secondary ip on a" |
672 |
" single-homed cluster requires the"
|
673 |
" --force option to be passed, and the"
|
674 |
" target node to be the master",
|
675 |
errors.ECODE_INVAL) |
676 |
elif not master_singlehomed and self.op.secondary_ip == node.primary_ip: |
677 |
if self.op.force and node.uuid == master.uuid: |
678 |
self.LogWarning("Transitioning from multi-homed to single-homed" |
679 |
" cluster; secondary IP addresses will have to be"
|
680 |
" removed")
|
681 |
else:
|
682 |
raise errors.OpPrereqError("Cannot set the secondary IP to be the" |
683 |
" same as the primary IP on a multi-homed"
|
684 |
" cluster, unless the --force option is"
|
685 |
" passed, and the target node is the"
|
686 |
" master", errors.ECODE_INVAL)
|
687 |
|
688 |
assert not (set([inst.name for inst in affected_instances.values()]) - |
689 |
self.owned_locks(locking.LEVEL_INSTANCE))
|
690 |
|
691 |
if node.offline:
|
692 |
if affected_instances:
|
693 |
msg = ("Cannot change secondary IP address: offline node has"
|
694 |
" instances (%s) configured to use it" %
|
695 |
utils.CommaJoin( |
696 |
[inst.name for inst in affected_instances.values()])) |
697 |
raise errors.OpPrereqError(msg, errors.ECODE_STATE)
|
698 |
else:
|
699 |
# On online nodes, check that no instances are running, and that
|
700 |
# the node has the new ip and we can reach it.
|
701 |
for instance in affected_instances.values(): |
702 |
CheckInstanceState(self, instance, INSTANCE_DOWN,
|
703 |
msg="cannot change secondary ip")
|
704 |
|
705 |
_CheckNodeHasSecondaryIP(self, node, self.op.secondary_ip, True) |
706 |
if master.uuid != node.uuid:
|
707 |
# check reachability from master secondary ip to new secondary ip
|
708 |
if not netutils.TcpPing(self.op.secondary_ip, |
709 |
constants.DEFAULT_NODED_PORT, |
710 |
source=master.secondary_ip): |
711 |
raise errors.OpPrereqError("Node secondary ip not reachable by TCP" |
712 |
" based ping to node daemon port",
|
713 |
errors.ECODE_ENVIRON) |
714 |
|
715 |
if self.op.ndparams: |
716 |
new_ndparams = GetUpdatedParams(node.ndparams, self.op.ndparams)
|
717 |
utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES) |
718 |
CheckParamsNotGlobal(self.op.ndparams, constants.NDC_GLOBALS, "node", |
719 |
"node", "cluster or group") |
720 |
self.new_ndparams = new_ndparams
|
721 |
|
722 |
if self.op.hv_state: |
723 |
self.new_hv_state = MergeAndVerifyHvState(self.op.hv_state, |
724 |
node.hv_state_static) |
725 |
|
726 |
if self.op.disk_state: |
727 |
self.new_disk_state = \
|
728 |
MergeAndVerifyDiskState(self.op.disk_state, node.disk_state_static)
|
729 |
|
730 |
def Exec(self, feedback_fn): |
731 |
"""Modifies a node.
|
732 |
|
733 |
"""
|
734 |
node = self.cfg.GetNodeInfo(self.op.node_uuid) |
735 |
result = [] |
736 |
|
737 |
if self.op.ndparams: |
738 |
node.ndparams = self.new_ndparams
|
739 |
|
740 |
if self.op.powered is not None: |
741 |
node.powered = self.op.powered
|
742 |
|
743 |
if self.op.hv_state: |
744 |
node.hv_state_static = self.new_hv_state
|
745 |
|
746 |
if self.op.disk_state: |
747 |
node.disk_state_static = self.new_disk_state
|
748 |
|
749 |
for attr in ["master_capable", "vm_capable"]: |
750 |
val = getattr(self.op, attr) |
751 |
if val is not None: |
752 |
setattr(node, attr, val)
|
753 |
result.append((attr, str(val)))
|
754 |
|
755 |
if self.new_role != self.old_role: |
756 |
# Tell the node to demote itself, if no longer MC and not offline
|
757 |
if self.old_role == self._ROLE_CANDIDATE and \ |
758 |
self.new_role != self._ROLE_OFFLINE: |
759 |
msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
|
760 |
if msg:
|
761 |
self.LogWarning("Node failed to demote itself: %s", msg) |
762 |
|
763 |
new_flags = self._R2F[self.new_role] |
764 |
for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS): |
765 |
if of != nf:
|
766 |
result.append((desc, str(nf)))
|
767 |
(node.master_candidate, node.drained, node.offline) = new_flags |
768 |
|
769 |
# we locked all nodes, we adjust the CP before updating this node
|
770 |
if self.lock_all: |
771 |
AdjustCandidatePool(self, [node.uuid])
|
772 |
|
773 |
if self.op.secondary_ip: |
774 |
node.secondary_ip = self.op.secondary_ip
|
775 |
result.append(("secondary_ip", self.op.secondary_ip)) |
776 |
|
777 |
# this will trigger configuration file update, if needed
|
778 |
self.cfg.Update(node, feedback_fn)
|
779 |
|
780 |
# this will trigger job queue propagation or cleanup if the mc
|
781 |
# flag changed
|
782 |
if [self.old_role, self.new_role].count(self._ROLE_CANDIDATE) == 1: |
783 |
self.context.ReaddNode(node)
|
784 |
|
785 |
return result
|
786 |
|
787 |
|
788 |
class LUNodePowercycle(NoHooksLU): |
789 |
"""Powercycles a node.
|
790 |
|
791 |
"""
|
792 |
REQ_BGL = False
|
793 |
|
794 |
def CheckArguments(self): |
795 |
(self.op.node_uuid, self.op.node_name) = \ |
796 |
ExpandNodeUuidAndName(self.cfg, self.op.node_uuid, self.op.node_name) |
797 |
|
798 |
if self.op.node_uuid == self.cfg.GetMasterNode() and not self.op.force: |
799 |
raise errors.OpPrereqError("The node is the master and the force" |
800 |
" parameter was not set",
|
801 |
errors.ECODE_INVAL) |
802 |
|
803 |
def ExpandNames(self): |
804 |
"""Locking for PowercycleNode.
|
805 |
|
806 |
This is a last-resort option and shouldn't block on other
|
807 |
jobs. Therefore, we grab no locks.
|
808 |
|
809 |
"""
|
810 |
self.needed_locks = {}
|
811 |
|
812 |
def Exec(self, feedback_fn): |
813 |
"""Reboots a node.
|
814 |
|
815 |
"""
|
816 |
default_hypervisor = self.cfg.GetHypervisorType()
|
817 |
hvparams = self.cfg.GetClusterInfo().hvparams[default_hypervisor]
|
818 |
result = self.rpc.call_node_powercycle(self.op.node_uuid, |
819 |
default_hypervisor, |
820 |
hvparams) |
821 |
result.Raise("Failed to schedule the reboot")
|
822 |
return result.payload
|
823 |
|
824 |
|
825 |
def _GetNodeInstancesInner(cfg, fn): |
826 |
return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)] |
827 |
|
828 |
|
829 |
def _GetNodePrimaryInstances(cfg, node_uuid): |
830 |
"""Returns primary instances on a node.
|
831 |
|
832 |
"""
|
833 |
return _GetNodeInstancesInner(cfg,
|
834 |
lambda inst: node_uuid == inst.primary_node)
|
835 |
|
836 |
|
837 |
def _GetNodeSecondaryInstances(cfg, node_uuid): |
838 |
"""Returns secondary instances on a node.
|
839 |
|
840 |
"""
|
841 |
return _GetNodeInstancesInner(cfg,
|
842 |
lambda inst: node_uuid in inst.secondary_nodes) |
843 |
|
844 |
|
845 |
def _GetNodeInstances(cfg, node_uuid): |
846 |
"""Returns a list of all primary and secondary instances on a node.
|
847 |
|
848 |
"""
|
849 |
|
850 |
return _GetNodeInstancesInner(cfg, lambda inst: node_uuid in inst.all_nodes) |
851 |
|
852 |
|
853 |
class LUNodeEvacuate(NoHooksLU): |
854 |
"""Evacuates instances off a list of nodes.
|
855 |
|
856 |
"""
|
857 |
REQ_BGL = False
|
858 |
|
859 |
def CheckArguments(self): |
860 |
CheckIAllocatorOrNode(self, "iallocator", "remote_node") |
861 |
|
862 |
def ExpandNames(self): |
863 |
(self.op.node_uuid, self.op.node_name) = \ |
864 |
ExpandNodeUuidAndName(self.cfg, self.op.node_uuid, self.op.node_name) |
865 |
|
866 |
if self.op.remote_node is not None: |
867 |
(self.op.remote_node_uuid, self.op.remote_node) = \ |
868 |
ExpandNodeUuidAndName(self.cfg, self.op.remote_node_uuid, |
869 |
self.op.remote_node)
|
870 |
assert self.op.remote_node |
871 |
|
872 |
if self.op.node_uuid == self.op.remote_node_uuid: |
873 |
raise errors.OpPrereqError("Can not use evacuated node as a new" |
874 |
" secondary node", errors.ECODE_INVAL)
|
875 |
|
876 |
if self.op.mode != constants.NODE_EVAC_SEC: |
877 |
raise errors.OpPrereqError("Without the use of an iallocator only" |
878 |
" secondary instances can be evacuated",
|
879 |
errors.ECODE_INVAL) |
880 |
|
881 |
# Declare locks
|
882 |
self.share_locks = ShareAll()
|
883 |
self.needed_locks = {
|
884 |
locking.LEVEL_INSTANCE: [], |
885 |
locking.LEVEL_NODEGROUP: [], |
886 |
locking.LEVEL_NODE: [], |
887 |
} |
888 |
|
889 |
# Determine nodes (via group) optimistically, needs verification once locks
|
890 |
# have been acquired
|
891 |
self.lock_nodes = self._DetermineNodes() |
892 |
|
893 |
def _DetermineNodes(self): |
894 |
"""Gets the list of node UUIDs to operate on.
|
895 |
|
896 |
"""
|
897 |
if self.op.remote_node is None: |
898 |
# Iallocator will choose any node(s) in the same group
|
899 |
group_nodes = self.cfg.GetNodeGroupMembersByNodes([self.op.node_uuid]) |
900 |
else:
|
901 |
group_nodes = frozenset([self.op.remote_node_uuid]) |
902 |
|
903 |
# Determine nodes to be locked
|
904 |
return set([self.op.node_uuid]) | group_nodes |
905 |
|
906 |
def _DetermineInstances(self): |
907 |
"""Builds list of instances to operate on.
|
908 |
|
909 |
"""
|
910 |
assert self.op.mode in constants.NODE_EVAC_MODES |
911 |
|
912 |
if self.op.mode == constants.NODE_EVAC_PRI: |
913 |
# Primary instances only
|
914 |
inst_fn = _GetNodePrimaryInstances |
915 |
assert self.op.remote_node is None, \ |
916 |
"Evacuating primary instances requires iallocator"
|
917 |
elif self.op.mode == constants.NODE_EVAC_SEC: |
918 |
# Secondary instances only
|
919 |
inst_fn = _GetNodeSecondaryInstances |
920 |
else:
|
921 |
# All instances
|
922 |
assert self.op.mode == constants.NODE_EVAC_ALL |
923 |
inst_fn = _GetNodeInstances |
924 |
# TODO: In 2.6, change the iallocator interface to take an evacuation mode
|
925 |
# per instance
|
926 |
raise errors.OpPrereqError("Due to an issue with the iallocator" |
927 |
" interface it is not possible to evacuate"
|
928 |
" all instances at once; specify explicitly"
|
929 |
" whether to evacuate primary or secondary"
|
930 |
" instances",
|
931 |
errors.ECODE_INVAL) |
932 |
|
933 |
return inst_fn(self.cfg, self.op.node_uuid) |
934 |
|
935 |
def DeclareLocks(self, level): |
936 |
if level == locking.LEVEL_INSTANCE:
|
937 |
# Lock instances optimistically, needs verification once node and group
|
938 |
# locks have been acquired
|
939 |
self.needed_locks[locking.LEVEL_INSTANCE] = \
|
940 |
set(i.name for i in self._DetermineInstances()) |
941 |
|
942 |
elif level == locking.LEVEL_NODEGROUP:
|
943 |
# Lock node groups for all potential target nodes optimistically, needs
|
944 |
# verification once nodes have been acquired
|
945 |
self.needed_locks[locking.LEVEL_NODEGROUP] = \
|
946 |
self.cfg.GetNodeGroupsFromNodes(self.lock_nodes) |
947 |
|
948 |
elif level == locking.LEVEL_NODE:
|
949 |
self.needed_locks[locking.LEVEL_NODE] = self.lock_nodes |
950 |
|
951 |
def CheckPrereq(self): |
952 |
# Verify locks
|
953 |
owned_instance_names = self.owned_locks(locking.LEVEL_INSTANCE)
|
954 |
owned_nodes = self.owned_locks(locking.LEVEL_NODE)
|
955 |
owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
|
956 |
|
957 |
need_nodes = self._DetermineNodes()
|
958 |
|
959 |
if not owned_nodes.issuperset(need_nodes): |
960 |
raise errors.OpPrereqError("Nodes in same group as '%s' changed since" |
961 |
" locks were acquired, current nodes are"
|
962 |
" are '%s', used to be '%s'; retry the"
|
963 |
" operation" %
|
964 |
(self.op.node_name,
|
965 |
utils.CommaJoin(need_nodes), |
966 |
utils.CommaJoin(owned_nodes)), |
967 |
errors.ECODE_STATE) |
968 |
|
969 |
wanted_groups = self.cfg.GetNodeGroupsFromNodes(owned_nodes)
|
970 |
if owned_groups != wanted_groups:
|
971 |
raise errors.OpExecError("Node groups changed since locks were acquired," |
972 |
" current groups are '%s', used to be '%s';"
|
973 |
" retry the operation" %
|
974 |
(utils.CommaJoin(wanted_groups), |
975 |
utils.CommaJoin(owned_groups))) |
976 |
|
977 |
# Determine affected instances
|
978 |
self.instances = self._DetermineInstances() |
979 |
self.instance_names = [i.name for i in self.instances] |
980 |
|
981 |
if set(self.instance_names) != owned_instance_names: |
982 |
raise errors.OpExecError("Instances on node '%s' changed since locks" |
983 |
" were acquired, current instances are '%s',"
|
984 |
" used to be '%s'; retry the operation" %
|
985 |
(self.op.node_name,
|
986 |
utils.CommaJoin(self.instance_names),
|
987 |
utils.CommaJoin(owned_instance_names))) |
988 |
|
989 |
if self.instance_names: |
990 |
self.LogInfo("Evacuating instances from node '%s': %s", |
991 |
self.op.node_name,
|
992 |
utils.CommaJoin(utils.NiceSort(self.instance_names)))
|
993 |
else:
|
994 |
self.LogInfo("No instances to evacuate from node '%s'", |
995 |
self.op.node_name)
|
996 |
|
997 |
if self.op.remote_node is not None: |
998 |
for i in self.instances: |
999 |
if i.primary_node == self.op.remote_node_uuid: |
1000 |
raise errors.OpPrereqError("Node %s is the primary node of" |
1001 |
" instance %s, cannot use it as"
|
1002 |
" secondary" %
|
1003 |
(self.op.remote_node, i.name),
|
1004 |
errors.ECODE_INVAL) |
1005 |
|
1006 |
def Exec(self, feedback_fn): |
1007 |
assert (self.op.iallocator is not None) ^ (self.op.remote_node is not None) |
1008 |
|
1009 |
if not self.instance_names: |
1010 |
# No instances to evacuate
|
1011 |
jobs = [] |
1012 |
|
1013 |
elif self.op.iallocator is not None: |
1014 |
# TODO: Implement relocation to other group
|
1015 |
req = iallocator.IAReqNodeEvac(evac_mode=self.op.mode,
|
1016 |
instances=list(self.instance_names)) |
1017 |
ial = iallocator.IAllocator(self.cfg, self.rpc, req) |
1018 |
|
1019 |
ial.Run(self.op.iallocator)
|
1020 |
|
1021 |
if not ial.success: |
1022 |
raise errors.OpPrereqError("Can't compute node evacuation using" |
1023 |
" iallocator '%s': %s" %
|
1024 |
(self.op.iallocator, ial.info),
|
1025 |
errors.ECODE_NORES) |
1026 |
|
1027 |
jobs = LoadNodeEvacResult(self, ial.result, self.op.early_release, True) |
1028 |
|
1029 |
elif self.op.remote_node is not None: |
1030 |
assert self.op.mode == constants.NODE_EVAC_SEC |
1031 |
jobs = [ |
1032 |
[opcodes.OpInstanceReplaceDisks(instance_name=instance_name, |
1033 |
remote_node=self.op.remote_node,
|
1034 |
disks=[], |
1035 |
mode=constants.REPLACE_DISK_CHG, |
1036 |
early_release=self.op.early_release)]
|
1037 |
for instance_name in self.instance_names] |
1038 |
|
1039 |
else:
|
1040 |
raise errors.ProgrammerError("No iallocator or remote node") |
1041 |
|
1042 |
return ResultWithJobs(jobs)
|
1043 |
|
1044 |
|
1045 |
class LUNodeMigrate(LogicalUnit): |
1046 |
"""Migrate all instances from a node.
|
1047 |
|
1048 |
"""
|
1049 |
HPATH = "node-migrate"
|
1050 |
HTYPE = constants.HTYPE_NODE |
1051 |
REQ_BGL = False
|
1052 |
|
1053 |
def CheckArguments(self): |
1054 |
pass
|
1055 |
|
1056 |
def ExpandNames(self): |
1057 |
(self.op.node_uuid, self.op.node_name) = \ |
1058 |
ExpandNodeUuidAndName(self.cfg, self.op.node_uuid, self.op.node_name) |
1059 |
|
1060 |
self.share_locks = ShareAll()
|
1061 |
self.needed_locks = {
|
1062 |
locking.LEVEL_NODE: [self.op.node_uuid],
|
1063 |
} |
1064 |
|
1065 |
def BuildHooksEnv(self): |
1066 |
"""Build hooks env.
|
1067 |
|
1068 |
This runs on the master, the primary and all the secondaries.
|
1069 |
|
1070 |
"""
|
1071 |
return {
|
1072 |
"NODE_NAME": self.op.node_name, |
1073 |
"ALLOW_RUNTIME_CHANGES": self.op.allow_runtime_changes, |
1074 |
} |
1075 |
|
1076 |
def BuildHooksNodes(self): |
1077 |
"""Build hooks nodes.
|
1078 |
|
1079 |
"""
|
1080 |
nl = [self.cfg.GetMasterNode()]
|
1081 |
return (nl, nl)
|
1082 |
|
1083 |
def CheckPrereq(self): |
1084 |
pass
|
1085 |
|
1086 |
def Exec(self, feedback_fn): |
1087 |
# Prepare jobs for migration instances
|
1088 |
jobs = [ |
1089 |
[opcodes.OpInstanceMigrate( |
1090 |
instance_name=inst.name, |
1091 |
mode=self.op.mode,
|
1092 |
live=self.op.live,
|
1093 |
iallocator=self.op.iallocator,
|
1094 |
target_node=self.op.target_node,
|
1095 |
allow_runtime_changes=self.op.allow_runtime_changes,
|
1096 |
ignore_ipolicy=self.op.ignore_ipolicy)]
|
1097 |
for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_uuid)] |
1098 |
|
1099 |
# TODO: Run iallocator in this opcode and pass correct placement options to
|
1100 |
# OpInstanceMigrate. Since other jobs can modify the cluster between
|
1101 |
# running the iallocator and the actual migration, a good consistency model
|
1102 |
# will have to be found.
|
1103 |
|
1104 |
assert (frozenset(self.owned_locks(locking.LEVEL_NODE)) == |
1105 |
frozenset([self.op.node_uuid])) |
1106 |
|
1107 |
return ResultWithJobs(jobs)
|
1108 |
|
1109 |
|
1110 |
def _GetStorageTypeArgs(cfg, storage_type): |
1111 |
"""Returns the arguments for a storage type.
|
1112 |
|
1113 |
"""
|
1114 |
# Special case for file storage
|
1115 |
|
1116 |
if storage_type == constants.ST_FILE:
|
1117 |
return [[cfg.GetFileStorageDir()]]
|
1118 |
elif storage_type == constants.ST_SHARED_FILE:
|
1119 |
dts = cfg.GetClusterInfo().enabled_disk_templates |
1120 |
paths = [] |
1121 |
if constants.DT_SHARED_FILE in dts: |
1122 |
paths.append(cfg.GetSharedFileStorageDir()) |
1123 |
if constants.DT_GLUSTER in dts: |
1124 |
paths.append(cfg.GetGlusterStorageDir()) |
1125 |
return [paths]
|
1126 |
else:
|
1127 |
return []
|
1128 |
|
1129 |
|
1130 |
class LUNodeModifyStorage(NoHooksLU): |
1131 |
"""Logical unit for modifying a storage volume on a node.
|
1132 |
|
1133 |
"""
|
1134 |
REQ_BGL = False
|
1135 |
|
1136 |
def CheckArguments(self): |
1137 |
(self.op.node_uuid, self.op.node_name) = \ |
1138 |
ExpandNodeUuidAndName(self.cfg, self.op.node_uuid, self.op.node_name) |
1139 |
|
1140 |
storage_type = self.op.storage_type
|
1141 |
|
1142 |
try:
|
1143 |
modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type] |
1144 |
except KeyError: |
1145 |
raise errors.OpPrereqError("Storage units of type '%s' can not be" |
1146 |
" modified" % storage_type,
|
1147 |
errors.ECODE_INVAL) |
1148 |
|
1149 |
diff = set(self.op.changes.keys()) - modifiable |
1150 |
if diff:
|
1151 |
raise errors.OpPrereqError("The following fields can not be modified for" |
1152 |
" storage units of type '%s': %r" %
|
1153 |
(storage_type, list(diff)),
|
1154 |
errors.ECODE_INVAL) |
1155 |
|
1156 |
def CheckPrereq(self): |
1157 |
"""Check prerequisites.
|
1158 |
|
1159 |
"""
|
1160 |
CheckStorageTypeEnabled(self.cfg.GetClusterInfo(), self.op.storage_type) |
1161 |
|
1162 |
def ExpandNames(self): |
1163 |
self.needed_locks = {
|
1164 |
locking.LEVEL_NODE: self.op.node_uuid,
|
1165 |
} |
1166 |
|
1167 |
def Exec(self, feedback_fn): |
1168 |
"""Computes the list of nodes and their attributes.
|
1169 |
|
1170 |
"""
|
1171 |
st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) |
1172 |
result = self.rpc.call_storage_modify(self.op.node_uuid, |
1173 |
self.op.storage_type, st_args,
|
1174 |
self.op.name, self.op.changes) |
1175 |
result.Raise("Failed to modify storage unit '%s' on %s" %
|
1176 |
(self.op.name, self.op.node_name)) |
1177 |
|
1178 |
|
1179 |
def _CheckOutputFields(fields, selected): |
1180 |
"""Checks whether all selected fields are valid according to fields.
|
1181 |
|
1182 |
@type fields: L{utils.FieldSet}
|
1183 |
@param fields: fields set
|
1184 |
@type selected: L{utils.FieldSet}
|
1185 |
@param selected: fields set
|
1186 |
|
1187 |
"""
|
1188 |
delta = fields.NonMatching(selected) |
1189 |
if delta:
|
1190 |
raise errors.OpPrereqError("Unknown output fields selected: %s" |
1191 |
% ",".join(delta), errors.ECODE_INVAL)
|
1192 |
|
1193 |
|
1194 |
class LUNodeQueryvols(NoHooksLU): |
1195 |
"""Logical unit for getting volumes on node(s).
|
1196 |
|
1197 |
"""
|
1198 |
REQ_BGL = False
|
1199 |
|
1200 |
def CheckArguments(self): |
1201 |
_CheckOutputFields(utils.FieldSet(constants.VF_NODE, constants.VF_PHYS, |
1202 |
constants.VF_VG, constants.VF_NAME, |
1203 |
constants.VF_SIZE, constants.VF_INSTANCE), |
1204 |
self.op.output_fields)
|
1205 |
|
1206 |
def ExpandNames(self): |
1207 |
self.share_locks = ShareAll()
|
1208 |
|
1209 |
if self.op.nodes: |
1210 |
self.needed_locks = {
|
1211 |
locking.LEVEL_NODE: GetWantedNodes(self, self.op.nodes)[0], |
1212 |
} |
1213 |
else:
|
1214 |
self.needed_locks = {
|
1215 |
locking.LEVEL_NODE: locking.ALL_SET, |
1216 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
1217 |
} |
1218 |
|
1219 |
def Exec(self, feedback_fn): |
1220 |
"""Computes the list of nodes and their attributes.
|
1221 |
|
1222 |
"""
|
1223 |
node_uuids = self.owned_locks(locking.LEVEL_NODE)
|
1224 |
volumes = self.rpc.call_node_volumes(node_uuids)
|
1225 |
|
1226 |
ilist = self.cfg.GetAllInstancesInfo()
|
1227 |
vol2inst = MapInstanceLvsToNodes(ilist.values()) |
1228 |
|
1229 |
output = [] |
1230 |
for node_uuid in node_uuids: |
1231 |
nresult = volumes[node_uuid] |
1232 |
if nresult.offline:
|
1233 |
continue
|
1234 |
msg = nresult.fail_msg |
1235 |
if msg:
|
1236 |
self.LogWarning("Can't compute volume data on node %s: %s", |
1237 |
self.cfg.GetNodeName(node_uuid), msg)
|
1238 |
continue
|
1239 |
|
1240 |
node_vols = sorted(nresult.payload,
|
1241 |
key=operator.itemgetter(constants.VF_DEV)) |
1242 |
|
1243 |
for vol in node_vols: |
1244 |
node_output = [] |
1245 |
for field in self.op.output_fields: |
1246 |
if field == constants.VF_NODE:
|
1247 |
val = self.cfg.GetNodeName(node_uuid)
|
1248 |
elif field == constants.VF_PHYS:
|
1249 |
val = vol[constants.VF_DEV] |
1250 |
elif field == constants.VF_VG:
|
1251 |
val = vol[constants.VF_VG] |
1252 |
elif field == constants.VF_NAME:
|
1253 |
val = vol[constants.VF_NAME] |
1254 |
elif field == constants.VF_SIZE:
|
1255 |
val = int(float(vol[constants.VF_SIZE])) |
1256 |
elif field == constants.VF_INSTANCE:
|
1257 |
inst = vol2inst.get((node_uuid, vol[constants.VF_VG] + "/" +
|
1258 |
vol[constants.VF_NAME]), None)
|
1259 |
if inst is not None: |
1260 |
val = inst.name |
1261 |
else:
|
1262 |
val = "-"
|
1263 |
else:
|
1264 |
raise errors.ParameterError(field)
|
1265 |
node_output.append(str(val))
|
1266 |
|
1267 |
output.append(node_output) |
1268 |
|
1269 |
return output
|
1270 |
|
1271 |
|
1272 |
class LUNodeQueryStorage(NoHooksLU): |
1273 |
"""Logical unit for getting information on storage units on node(s).
|
1274 |
|
1275 |
"""
|
1276 |
REQ_BGL = False
|
1277 |
|
1278 |
def CheckArguments(self): |
1279 |
_CheckOutputFields(utils.FieldSet(*constants.VALID_STORAGE_FIELDS), |
1280 |
self.op.output_fields)
|
1281 |
|
1282 |
def ExpandNames(self): |
1283 |
self.share_locks = ShareAll()
|
1284 |
|
1285 |
if self.op.nodes: |
1286 |
self.needed_locks = {
|
1287 |
locking.LEVEL_NODE: GetWantedNodes(self, self.op.nodes)[0], |
1288 |
} |
1289 |
else:
|
1290 |
self.needed_locks = {
|
1291 |
locking.LEVEL_NODE: locking.ALL_SET, |
1292 |
locking.LEVEL_NODE_ALLOC: locking.ALL_SET, |
1293 |
} |
1294 |
|
1295 |
def _DetermineStorageType(self): |
1296 |
"""Determines the default storage type of the cluster.
|
1297 |
|
1298 |
"""
|
1299 |
enabled_disk_templates = self.cfg.GetClusterInfo().enabled_disk_templates
|
1300 |
default_storage_type = \ |
1301 |
constants.MAP_DISK_TEMPLATE_STORAGE_TYPE[enabled_disk_templates[0]]
|
1302 |
return default_storage_type
|
1303 |
|
1304 |
def CheckPrereq(self): |
1305 |
"""Check prerequisites.
|
1306 |
|
1307 |
"""
|
1308 |
if self.op.storage_type: |
1309 |
CheckStorageTypeEnabled(self.cfg.GetClusterInfo(), self.op.storage_type) |
1310 |
self.storage_type = self.op.storage_type |
1311 |
else:
|
1312 |
self.storage_type = self._DetermineStorageType() |
1313 |
supported_storage_types = constants.STS_REPORT_NODE_STORAGE |
1314 |
if self.storage_type not in supported_storage_types: |
1315 |
raise errors.OpPrereqError(
|
1316 |
"Storage reporting for storage type '%s' is not supported. Please"
|
1317 |
" use the --storage-type option to specify one of the supported"
|
1318 |
" storage types (%s) or set the default disk template to one that"
|
1319 |
" supports storage reporting." %
|
1320 |
(self.storage_type, utils.CommaJoin(supported_storage_types)))
|
1321 |
|
1322 |
def Exec(self, feedback_fn): |
1323 |
"""Computes the list of nodes and their attributes.
|
1324 |
|
1325 |
"""
|
1326 |
if self.op.storage_type: |
1327 |
self.storage_type = self.op.storage_type |
1328 |
else:
|
1329 |
self.storage_type = self._DetermineStorageType() |
1330 |
|
1331 |
self.node_uuids = self.owned_locks(locking.LEVEL_NODE) |
1332 |
|
1333 |
# Always get name to sort by
|
1334 |
if constants.SF_NAME in self.op.output_fields: |
1335 |
fields = self.op.output_fields[:]
|
1336 |
else:
|
1337 |
fields = [constants.SF_NAME] + self.op.output_fields
|
1338 |
|
1339 |
# Never ask for node or type as it's only known to the LU
|
1340 |
for extra in [constants.SF_NODE, constants.SF_TYPE]: |
1341 |
while extra in fields: |
1342 |
fields.remove(extra) |
1343 |
|
1344 |
field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)]) |
1345 |
name_idx = field_idx[constants.SF_NAME] |
1346 |
|
1347 |
st_args = _GetStorageTypeArgs(self.cfg, self.storage_type) |
1348 |
data = self.rpc.call_storage_list(self.node_uuids, |
1349 |
self.storage_type, st_args,
|
1350 |
self.op.name, fields)
|
1351 |
|
1352 |
result = [] |
1353 |
|
1354 |
for node_uuid in utils.NiceSort(self.node_uuids): |
1355 |
node_name = self.cfg.GetNodeName(node_uuid)
|
1356 |
nresult = data[node_uuid] |
1357 |
if nresult.offline:
|
1358 |
continue
|
1359 |
|
1360 |
msg = nresult.fail_msg |
1361 |
if msg:
|
1362 |
self.LogWarning("Can't get storage data from node %s: %s", |
1363 |
node_name, msg) |
1364 |
continue
|
1365 |
|
1366 |
rows = dict([(row[name_idx], row) for row in nresult.payload]) |
1367 |
|
1368 |
for name in utils.NiceSort(rows.keys()): |
1369 |
row = rows[name] |
1370 |
|
1371 |
out = [] |
1372 |
|
1373 |
for field in self.op.output_fields: |
1374 |
if field == constants.SF_NODE:
|
1375 |
val = node_name |
1376 |
elif field == constants.SF_TYPE:
|
1377 |
val = self.storage_type
|
1378 |
elif field in field_idx: |
1379 |
val = row[field_idx[field]] |
1380 |
else:
|
1381 |
raise errors.ParameterError(field)
|
1382 |
|
1383 |
out.append(val) |
1384 |
|
1385 |
result.append(out) |
1386 |
|
1387 |
return result
|
1388 |
|
1389 |
|
1390 |
class LUNodeRemove(LogicalUnit): |
1391 |
"""Logical unit for removing a node.
|
1392 |
|
1393 |
"""
|
1394 |
HPATH = "node-remove"
|
1395 |
HTYPE = constants.HTYPE_NODE |
1396 |
|
1397 |
def BuildHooksEnv(self): |
1398 |
"""Build hooks env.
|
1399 |
|
1400 |
"""
|
1401 |
return {
|
1402 |
"OP_TARGET": self.op.node_name, |
1403 |
"NODE_NAME": self.op.node_name, |
1404 |
} |
1405 |
|
1406 |
def BuildHooksNodes(self): |
1407 |
"""Build hooks nodes.
|
1408 |
|
1409 |
This doesn't run on the target node in the pre phase as a failed
|
1410 |
node would then be impossible to remove.
|
1411 |
|
1412 |
"""
|
1413 |
all_nodes = self.cfg.GetNodeList()
|
1414 |
try:
|
1415 |
all_nodes.remove(self.op.node_uuid)
|
1416 |
except ValueError: |
1417 |
pass
|
1418 |
return (all_nodes, all_nodes)
|
1419 |
|
1420 |
def CheckPrereq(self): |
1421 |
"""Check prerequisites.
|
1422 |
|
1423 |
This checks:
|
1424 |
- the node exists in the configuration
|
1425 |
- it does not have primary or secondary instances
|
1426 |
- it's not the master
|
1427 |
|
1428 |
Any errors are signaled by raising errors.OpPrereqError.
|
1429 |
|
1430 |
"""
|
1431 |
(self.op.node_uuid, self.op.node_name) = \ |
1432 |
ExpandNodeUuidAndName(self.cfg, self.op.node_uuid, self.op.node_name) |
1433 |
node = self.cfg.GetNodeInfo(self.op.node_uuid) |
1434 |
assert node is not None |
1435 |
|
1436 |
masternode = self.cfg.GetMasterNode()
|
1437 |
if node.uuid == masternode:
|
1438 |
raise errors.OpPrereqError("Node is the master node, failover to another" |
1439 |
" node is required", errors.ECODE_INVAL)
|
1440 |
|
1441 |
for _, instance in self.cfg.GetAllInstancesInfo().items(): |
1442 |
if node.uuid in instance.all_nodes: |
1443 |
raise errors.OpPrereqError("Instance %s is still running on the node," |
1444 |
" please remove first" % instance.name,
|
1445 |
errors.ECODE_INVAL) |
1446 |
self.op.node_name = node.name
|
1447 |
self.node = node
|
1448 |
|
1449 |
def Exec(self, feedback_fn): |
1450 |
"""Removes the node from the cluster.
|
1451 |
|
1452 |
"""
|
1453 |
logging.info("Stopping the node daemon and removing configs from node %s",
|
1454 |
self.node.name)
|
1455 |
|
1456 |
modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
|
1457 |
|
1458 |
assert locking.BGL in self.owned_locks(locking.LEVEL_CLUSTER), \ |
1459 |
"Not owning BGL"
|
1460 |
|
1461 |
# Promote nodes to master candidate as needed
|
1462 |
AdjustCandidatePool(self, exceptions=[self.node.uuid]) |
1463 |
self.context.RemoveNode(self.node) |
1464 |
|
1465 |
# Run post hooks on the node before it's removed
|
1466 |
RunPostHook(self, self.node.name) |
1467 |
|
1468 |
# we have to call this by name rather than by UUID, as the node is no longer
|
1469 |
# in the config
|
1470 |
result = self.rpc.call_node_leave_cluster(self.node.name, modify_ssh_setup) |
1471 |
msg = result.fail_msg |
1472 |
if msg:
|
1473 |
self.LogWarning("Errors encountered on the remote node while leaving" |
1474 |
" the cluster: %s", msg)
|
1475 |
|
1476 |
# Remove node from our /etc/hosts
|
1477 |
if self.cfg.GetClusterInfo().modify_etc_hosts: |
1478 |
master_node_uuid = self.cfg.GetMasterNode()
|
1479 |
result = self.rpc.call_etc_hosts_modify(master_node_uuid,
|
1480 |
constants.ETC_HOSTS_REMOVE, |
1481 |
self.node.name, None) |
1482 |
result.Raise("Can't update hosts file with new host data")
|
1483 |
RedistributeAncillaryFiles(self)
|
1484 |
|
1485 |
|
1486 |
class LURepairNodeStorage(NoHooksLU): |
1487 |
"""Repairs the volume group on a node.
|
1488 |
|
1489 |
"""
|
1490 |
REQ_BGL = False
|
1491 |
|
1492 |
def CheckArguments(self): |
1493 |
(self.op.node_uuid, self.op.node_name) = \ |
1494 |
ExpandNodeUuidAndName(self.cfg, self.op.node_uuid, self.op.node_name) |
1495 |
|
1496 |
storage_type = self.op.storage_type
|
1497 |
|
1498 |
if (constants.SO_FIX_CONSISTENCY not in |
1499 |
constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])): |
1500 |
raise errors.OpPrereqError("Storage units of type '%s' can not be" |
1501 |
" repaired" % storage_type,
|
1502 |
errors.ECODE_INVAL) |
1503 |
|
1504 |
def ExpandNames(self): |
1505 |
self.needed_locks = {
|
1506 |
locking.LEVEL_NODE: [self.op.node_uuid],
|
1507 |
} |
1508 |
|
1509 |
def _CheckFaultyDisks(self, instance, node_uuid): |
1510 |
"""Ensure faulty disks abort the opcode or at least warn."""
|
1511 |
try:
|
1512 |
if FindFaultyInstanceDisks(self.cfg, self.rpc, instance, |
1513 |
node_uuid, True):
|
1514 |
raise errors.OpPrereqError("Instance '%s' has faulty disks on" |
1515 |
" node '%s'" %
|
1516 |
(instance.name, |
1517 |
self.cfg.GetNodeName(node_uuid)),
|
1518 |
errors.ECODE_STATE) |
1519 |
except errors.OpPrereqError, err:
|
1520 |
if self.op.ignore_consistency: |
1521 |
self.LogWarning(str(err.args[0])) |
1522 |
else:
|
1523 |
raise
|
1524 |
|
1525 |
def CheckPrereq(self): |
1526 |
"""Check prerequisites.
|
1527 |
|
1528 |
"""
|
1529 |
CheckStorageTypeEnabled(self.cfg.GetClusterInfo(), self.op.storage_type) |
1530 |
|
1531 |
# Check whether any instance on this node has faulty disks
|
1532 |
for inst in _GetNodeInstances(self.cfg, self.op.node_uuid): |
1533 |
if not inst.disks_active: |
1534 |
continue
|
1535 |
check_nodes = set(inst.all_nodes)
|
1536 |
check_nodes.discard(self.op.node_uuid)
|
1537 |
for inst_node_uuid in check_nodes: |
1538 |
self._CheckFaultyDisks(inst, inst_node_uuid)
|
1539 |
|
1540 |
def Exec(self, feedback_fn): |
1541 |
feedback_fn("Repairing storage unit '%s' on %s ..." %
|
1542 |
(self.op.name, self.op.node_name)) |
1543 |
|
1544 |
st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type) |
1545 |
result = self.rpc.call_storage_execute(self.op.node_uuid, |
1546 |
self.op.storage_type, st_args,
|
1547 |
self.op.name,
|
1548 |
constants.SO_FIX_CONSISTENCY) |
1549 |
result.Raise("Failed to repair storage unit '%s' on %s" %
|
1550 |
(self.op.name, self.op.node_name)) |